Peripheral and tumor‐infiltrating immune cells are correlated with patient outcomes in ovarian cancer

Weiwei Zhang, Yawen Ling, Zhidong Li, Xingchen Peng, Yazhou RenCancer 2023 · DOI

TL;DR

This study leverages machine learning to identify peripheral blood markers and tumor-infiltrating immune cells (TICs) that predict ovarian cancer (OC) patient survival. They found that specific immune-related genes (CXCL9, CD79A, MS4A1, MZB1) are strongly associated with both TICs and patient outcomes, suggesting their potential as prognostic biomarkers and therapeutic targets for OC immunotherapy.

What Problem Does This Paper Address?

Ovarian cancer (OC) remains a deadly malignancy with high recurrence rates and limited effective treatment options. While immunotherapy has shown promise in various cancers, its efficacy in OC is moderate, partly due to a lack of reliable biomarkers to predict prognosis and identify patients who would benefit most. Current biomarkers like PD-L1 and tumor mutational burden have limitations, including invasiveness and variability. This paper aims to address this gap by exploring the relationship between easily accessible peripheral blood indicators and the complex tumor immune microenvironment to find novel, cost-effective prognostic markers for OC.

Methods

The researchers employed a multi-pronged approach:

  • Machine Learning for Prognostic Prediction: Four machine learning algorithms (SVM, RF, XGBoost, and LightGBM) were used to predict overall survival (OS) in OC patients based on peripheral blood clinical indicators. LightGBM demonstrated superior performance.
  • TIC Analysis: The CIBERSORT algorithm was used to quantify the proportion of 22 different TICs within tumor samples from the TCGA database.
  • Immune Microenvironment Clustering: Consensus clustering was applied to group OC patients based on their TIC profiles, revealing two distinct groups (A and B) with significantly different survival rates.
  • Differential Gene Expression Analysis: Differentially expressed genes (DEGs) between the two TIC groups were identified.
  • Network and Pathway Analysis:
    • A protein-protein interaction (PPI) network was constructed for the DEGs.
  • Univariate COX regression analysis was performed to identify DEGs significantly associated with patient prognosis.
  • The intersection of DEGs from the PPI network and COX regression identified four core immune-related genes.
  • Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were conducted on DEGs to understand their functional pathways.
  • Gene Set Enrichment Analysis (GSEA) was used to explore the signaling pathways associated with the four core genes.

Datasets

  • Peripheral Blood Clinical Indicators: Data from 342 OC patients treated at the Fifth People's Hospital of Chengdu (2005-2015) were used. This included age, stage, chemoresistance, CA-125, surgical satisfaction, ascites, histological grade/type, CRP, PLT, LC, NEU, ALB, HGB, BASO, Mono, WBC, BMI, PLR, LMR, and NLR.
  • Tumor Immune Microenvironment Data: Transcriptome RNA-seq data (HTSeq-FPKM) and clinical data for 379 OC patients were downloaded from The Cancer Genome Atlas (TCGA) database (TCGA-OV).

Key Insights and Ideas

  • Peripheral Blood as a Prognostic Indicator: The study strongly suggests that readily available peripheral blood indicators are significant prognostic factors for OC patients, a finding supported by the performance of machine learning models.
  • TICs Drive Survival Differences: Clustering OC patients based on their TIC profiles revealed distinct groups with significantly different survival outcomes, highlighting the critical role of the immune microenvironment in OC progression.
  • Identification of Core Immune Genes: The integration of PPI network analysis and univariate COX regression successfully pinpointed four key immune-related genes (CXCL9, CD79A, MS4A1, and MZB1) that are significantly associated with both TICs and patient survival.
  • Therapeutic Potential of Core Genes: The strong correlation of these four genes with immune cells and their enrichment in immune-related pathways suggest they could serve as novel therapeutic targets for OC immunotherapy.

Key Results

  • Machine Learning Performance: LightGBM achieved the highest accuracy (0.88) in predicting OC patient survival based on peripheral blood indicators, demonstrating the power of machine learning in this context.
  • Important Peripheral Blood Indicators: Feature importance analysis identified tumor stage, histological grade, PLT, CRP, LMR, and HGB as key peripheral blood indicators influencing OC survival.
  • TICs and Survival: Kaplan-Meier survival analysis showed a significant association between TICs and OC patient survival, with one group exhibiting a higher survival rate.
  • Four Core Genes Identified: CXCL9, CD79A, MS4A1, and MZB1 were identified as core genes significantly associated with both OC survival and TICs.
  • Gene Expression and Survival: High expression of these four genes was consistently associated with better patient prognosis, while low expression indicated a poorer prognosis.
  • Immune Pathway Enrichment: GSEA confirmed that these four genes are involved in crucial immune-related signaling pathways, including those related to B cells, T cells, and macrophages.

Limitations and Future Work

The authors acknowledge several limitations:

  • Retrospective Nature: The study relies on retrospective data, and the findings need validation through prospective, multicenter clinical trials.
  • Data Integration Challenges: The TCGA database lacks peripheral blood indicators, and the hospital data lacks TICs and genetic information, preventing a direct correlation analysis between peripheral blood and TICs within the same dataset.
  • Machine Learning Familiarity: The limited familiarity of clinicians with machine learning methods presents a barrier to the widespread adoption of these tools.

Future work will focus on:

  • Prospective Clinical Studies: Conducting multicenter, multi-sample prospective studies to validate the findings and investigate the guiding role of combining clinical blood indicators and TICs.
  • Developing User-Friendly Tools: Creating visualized machine-learning methods or APP-like software to make these powerful analytical tools more accessible to clinicians.
  • Further Investigating Core Genes: Deeper studies into the identified core genes (CXCL9, CD79A, MS4A1, MZB1) to explore their precise mechanisms and potential as therapeutic targets or biomarkers for OC.

Summary updated

Abstract

These results suggest that both peripheral blood markers and TICs can be used as prognostic predictors in patients with OC, and CXCL9, CD79A, MS4A1, and MZB1 may be potential therapeutic targets for OC immunotherapy.
BibTeX
@inproceedings{zhang2023peripheral,
  author = {Weiwei Zhang and Yawen Ling and Zhidong Li and Xingchen Peng and Yazhou Ren},
  title = {{Peripheral and tumor‐infiltrating immune cells are correlated with patient outcomes in ovarian cancer}},
  booktitle = {Cancer},
  year = {2023},
  doi = {10.1002/cam4.5590},
}
Figure